A consideration of errors and accuracy in the isoconversional methods

A consideration of errors and accuracy in the isoconversional methods

Thermochimica Acta 369 (2001) 137±142 A consideration of errors and accuracy in the isoconversional methods Zhiming Gaoa,*, Masahiro Nakadaa, Iwao Am...

100KB Sizes 1 Downloads 180 Views

Thermochimica Acta 369 (2001) 137±142

A consideration of errors and accuracy in the isoconversional methods Zhiming Gaoa,*, Masahiro Nakadaa, Iwao Amasakib a

Department of Chemistry, Chiba Institute of Technology, 2-1-1 Shibazono, Narashino-shi, Chiba 275-0023, Japan Department of Electrical Engineering, Chiba Institute of Technology, 2-17-1 Tsudanuma, Narashino-shi, Chiba 275-0016, Japan

b

Received 6 April 2000; received in revised form 28 July 2000; accepted 27 November 2000

Abstract The conventional isoconversional plots, ln b versus 1/T and ln(b/T2) versus 1/T, have some errors in calculating the activation energy due to the use of an approximate value of p(x). In order to obtain the `exact' value of activation energy of a reaction, an iterative procedure has been established after the ratios, H(x) and h(x), were introduced to determine `exact' p(x) instead of the approximate values of p(x) used in the conventional isoconversional plots. The iterative procedure is wide applicable, no matter how little or how great the E/RT value of the reaction is. # 2001 Elsevier Science B.V. All rights reserved. Keywords: Iterative procedure; Activation energy; Isoconversional methods; Non-isothermal; Kinetics

1. Introduction The isoconversional methods, Ozawa method [1] and Kissinger±Akahira±Sunose method [2,3], have been widely used to estimate activation energies regardless of the rate expressions of the reactions. The successful application is ascribed to the approximate expressions used in the integration of Arrhenius equation [4,5]. For the same reason, however, the calculated activation energies contain some errors, which are dependent on E/RT value. An analysis of the error of Kissinger±Akahira±Sunose method was reported by Ortega et al. [6]. In their work, the error was evaluated at separate E/RT values in comparison to the `exact' integral of the Arrhenius equation. In practice, calculation of activation energy in the isoconversional methods is made over a range of E/RT values corresponding * Corresponding author. Fax: ‡81-47-454-9659. E-mail address: [email protected] (Z. Gao).

to the temperatures at different heating rates. The error of calculated activation energy is thus concerned in a regressive way with all the errors at the E/RT values within the range. This is the reason why Ozawa method can give a useful estimate of the activation energy usually although the approximation adopted in integrating the Arrhenius equation is rather rough. Nevertheless, the errors in the calculated activation energy will become larger as E/RT value decreases below 15. In the present work, through a thorough consideration of the errors of the conventional isoconversional methods, we present an improved scheme to obtain exact values for the activation energy, no matter how little or how great the E/RT value of the reaction is. 2. Theoretical In non-isothermal analyses, the reaction rate is generally expressed by the derivative of the fractional

0040-6031/01/$ ± see front matter # 2001 Elsevier Science B.V. All rights reserved. PII: S 0 0 4 0 - 6 0 3 1 ( 0 0 ) 0 0 7 6 0 - 7

138

Z. Gao et al. / Thermochimica Acta 369 (2001) 137±142

strictest sense.

conversion as Eq. (1). da ˆ AeÿE=RT f …a† dt Furthermore Z t Z a da ˆ AeÿE=RT dt g…a† ˆ f …a† 0 0 Z AE 1 exp…ÿx† AE p…x† dx ˆ ˆ bR x x2 bR

(1)

p…x† ˆ

exp…ÿx† x2

(2)

(3) (4)

From Eq. (3) Ozawa adopted a linear plot of ln b against 1/T at the same fractional conversion. Similarly, a linear plot of ln(b/T2) against 1/T from Eq. (4) was adopted by Kissinger et al.   0:0048AE 1:0516E ÿ ln g…a† ÿ (5) ln b ˆ ln R RT   b AR E ÿ ln g…a† ÿ (6) ln 2 ˆ ln T E RT These two methods of plotting a linear regressive curve at the same fractional conversion are the socalled isoconversional methods. The plot of ln(b/T2) against 1/T in Eq. (6) has been proved to give an activation energy with an error within 1% when x is greater than 15 compared with the use of the `exact' integral of Arrhenius equation [6]. The `exact' p(x) was regarded as Eq. (7). p…x† ˆ

exp…ÿx† h…x† x2

(7)

where h…x† ˆ

x4

x4 ‡ 18x3 ‡ 88x2 ‡ 96x ‡ 20x3 ‡ 120x2 ‡ 240x ‡ 120

dm ˆ AeÿE=RT f …m† dt

(9)

Of course, because m ˆ m0 …1 ÿ a†, Eq. (9) can be rewritten as

where x ˆ E=RT, T0 ˆ 0 K, b is the heating rate of a linear temperature-programmed process. As is well known, the p function has no analytical solution and is expressed by several approximate expressions [4,5]. p…x† ˆ 0:0048 exp…ÿ1:0516x†

ÿ

(8)

As a matter of fact, reaction rate should be expressed by derivative of mass decrease of a reactant (or mass increase of reaction product) as Eq. (9) in the

da 1 ˆ AeÿE=RT f …m0 ; a† dt m0

(10)

When the reaction under study is a nth order reaction, m0 can be separated from a, that is da n ˆ AeÿE=RT mnÿ1 0 …1 ÿ a† dt

(11)

In this way, Eqs. (5) and (6) should be rewritten as Eqs. (12) and (13).   0:0048AE ÿ ln g…m0 ; a† ÿ ln m0 ln b ˆ ln R 1:0516E (12) ÿ RT   b AR E ÿ ln g…m0 ; a† ÿ ln m0 ÿ (13) ln 2 ˆ ln T E RT It is evident that the initial mass of sample m0 must be kept at constant in all measurements at different heating rates to maintain a rigorous linear relationship for Eqs. (12) and (13). However, it is often seen that the initial mass of sample was not kept constant in the literature [7,8]. When m0 is separated with a in f(m0, a), keeping the same initial mass in all measurements is not necessary since the term of m0 can be combined into the term containing b. If it is supposed that the initial mass of sample is kept constant in all measurements at different heating rates, then Eqs. (5) and (6) can be adopted in the following discussion for convenience. Here, we also regard Eq. (7) as exact p(x), then evaluate the errors of activation energies obtained from Eqs. (5) and (6) in comparison to the exact p(x). It is certain that more accurate the used approximate expression of p(x), more accurate the calculated activation energies. Even although, if there is an approximate p(x), no matter how rough it is, having a constant ratio to the exact p(x), the approximate p(x) will give the same value of activation energy as the exact p(x) does. The ratio of the exact p(x) (Eq. (7)) to Eq. (4) is h(x) (Eq. (8)), which varies slightly with x, from 0.8910 to 0.9729

Z. Gao et al. / Thermochimica Acta 369 (2001) 137±142

139

Fig. 1. The variations of H(x) (open circle) and h(x) (black circle) with x.

corresponding to x ˆ 15±70. The ratio of the exact p(x) to Eq. (3) is de®ned as H(x). H…x† ˆ

exp…ÿx†h…x†=x2 0:0048 exp…ÿ1:0516x†

(14)

H(x) was found to vary greatly with x value. The variations of H(x) and h(x) with x are shown in Fig. 1. The form of the function H(x) can be obtained by ®tting the ratios against x, but it has to be a long polynomial to re¯ect accurately these variations. After taking h(x) and H(x) into account, Eqs. (5) and (6) can be rewritten as Eqs. (15) and (16).   b 0:0048AE 1:0516E ÿ ln g…a† ÿ (15) ln ˆ ln H R RT   b AR E ÿ ln g…a† ÿ (16) ln 2 ˆ ln hT E RT Eqs. (15) and (16) are perfect relations since the exact function of p(x) is used. It can be seen that plots of ln b versus 1/T and ln(b/T2) versus 1/T in the conventional isoconversional methods are consequences of ignoring the variation of H(x) and h(x) with x. This is the reason for errors in the conventional isoconversional methods. In an accurate kinetic study, an iterative calculation should be performed by the use

of Eq. (15) or Eq. (16). The iterative procedure is as follows.  Step 1: supposing H…x† ˆ 1 or h…x† ˆ 1 to estimate the initial value of the activation energy E1. The conventional isoconversional methods stop calculating at this step.  Step 2: using E1 calculate H(x) or h(x), then from Eq. (15) or Eq. (16) calculate a new value E2 for the activation energy from the plot of ln(b/H) versus 1/ T or ln(b/hT2) versus 1/T.  Step 3: repeat step 2, replacing E1 with E2. And so on until the absolute difference of (Ei ÿ Eiÿ1 ) is less than a de®ned small quantity such as 0.1 kJ molÿ1 generally. The last value Ei is the exact value of activation energy of the reaction. The error in the conventional isoconversional calculation of activation energy can then be given by value of (E1 ÿ Ei )/ Ei. If the reaction mechanism is known, the exact preexponential factor A can be calculated from the intercept of the plot at the same time. On the other hand, one can ®t an approximate p(x) more accurate than Eq. (3) over the range of x of the being considered reaction for evaluating error of activation energy calculated from Ozawa plot with the use of Eq. (3). As ordinary reactions correspond to

140

Z. Gao et al. / Thermochimica Acta 369 (2001) 137±142

Table 1 The ®tted p(x) at the intervals of Dx ˆ 5 from the values of the exact p(x) vs. x x

The fitted p(x)

10±15 15±20 20±25 25±30 30±35 35±40 40±45 45±50 50±55 55±60 60±65 65±70

0.0376 exp(ÿ1.1516x) 0.0202 exp(ÿ1.1095x) 0.0126 exp(ÿ1.0858x) 0.0086 exp(ÿ1.0706x) 0.0063 exp(ÿ1.0600x) 0.0048 exp(ÿ1.0521x) 0.0038 exp(ÿ1.0461x) 0.0030 exp(ÿ1.0413x) 0.0025 exp(ÿ1.0375x) 0.0021 exp(ÿ1.0342x) 0.0018 exp(ÿ1.0315x) 0.0015 exp(ÿ1.0292x)

range of Dx around 3±4 at usual heating rates, the ®tting of a more accurate approximate p(x) in the form of Eq. (3) from the exact p(x) (Eq. (7)) was made at several intervals of Dx ˆ 5. The approximate p(x) function are indicated in Table 1. It can be seen that Eq. (3) is very similar to the ®tted p(x) at x ˆ 35±40. So it is obvious that Ozawa plot using Eq. (3) will lead to a larger value at lower x values and a less value at higher x values than the exact values of activation energy. 3. Experimental Thermal decomposition of a polyamide 6 (lot No. 1013B, Ube Industries Ltd., Japan) was carried out in a thermogravimetric analyzer (Shimadzu, TGA-50, Japan) in a ¯owing atmosphere of dry nitrogen (¯ow rate 40 cm3 minÿ1) from 373 to 873 K at heating rate of 2, 4, 6 and 10 K minÿ1, respectively. The sample cells were platinum crucibles. The sample mass was 4.867±5.384 mg. Calcium oxalate monohydrate (of special grade, Kanto Chemical Co., Japan) was dehy-

drated in the same thermogravimetric analyzer in nitrogen gas ¯ow of 40 cm3 minÿ1 from ambient temperature at heating rate of 1, 2, 5 and 10 K minÿ1, respectively. The sample mass was 3.687±7.110 mg. 4. Results and discussion It is reported that the polyamide 6 decomposes by a ®rst-order reaction [9,10] with the formation of the charred residue about 4 wt.% at the end [11]. So the initial mass of the sample should have no effect on the calculation of the activation energy for polyamide 6 according to Eq. (11) except the last part of the decomposition. To illustrate the iterative method, iterative calculations of the activation energy for the thermal decomposition of the polyamide 6 were carried out at a ˆ 0:3. The results are indicated in Tables 2 and 3. It can be seen from Tables 2 and 3 that the same value of activation energy was obtained at their last step of iterative calculations. This is because the same `exact' p(x) (Eq. (7)) was used in both Eqs. (15) and (16). But Eq. (5) contained a larger error than Eq. (6) in the initial activation energy. This can be understood from Fig. 1. Here, h(x) is slowly varying with x, whereas H(x) varies greatly with a small change of x except in the range of x ˆ 35±40. So Eq. (4) is better for calculating the activation energy. Similar results were seen in the iterative calculations on the dehydration of calcium oxalate monohydrate as shown in Tables 4 and 5. The effect of the initial mass of the sample was not considered as the mechanism of the dehydration was not known. Discussion of the mechanism is beyond the scope of this paper. In this case Ozawa plot exhibited a slightly larger error (2.6%) in the activation energy because the E/RT values were around 24, lower than those (around 34) for the thermal decomposition of the polyamide 6. According to Fig. 1, the Ozawa plot

Table 2 An iterative calculation of activation energy for the thermal decomposition of the polyamide 6 by use of the plot of ln(b/H) vs. 1/T at a ˆ 0.3 b (K minÿ1)

T (K)

h1(x)

E1 (kJ molÿ1)

h2(x)

E2 (kJ molÿ1)

h3(x)

E3 (kJ molÿ1)

2 4 6 10

684.8 697.4 706.6 716.0

1.0 1.0 1.0 1.0

197.6

0.9827 0.9858 0.9884 0.9914

196.6

0.9836 0.9869 0.9896 0.9927

196.5

Z. Gao et al. / Thermochimica Acta 369 (2001) 137±142

141

Table 3 An iterative calculation of activation energy for the thermal decomposition of the polyamide 6 by use of the plot of ln(b/hT2) vs. 1/T at a ˆ 0.3 b (K minÿ1)

T (K)

h1(x)

E1 (kJ molÿ1)

h2(x)

E2 (kJ molÿ1)

h3(x)

E3 (kJ molÿ1)

2 4 6 10

684.8 697.4 706.6 716.0

1.0 1.0 1.0 1.0

196.2

0.9475 0.9466 0.9460 0.9453

196.5

0.9476 0.9467 0.9461 0.9454

196.5

Table 4 An iterative calculation of activation energy for dehydration of the calcium oxalate monohydrate by use of the plot of ln(b/H) vs. 1/T at a ˆ 0.4 b (K minÿ1)

T (K)

h1(x)

E1 (kJ molÿ1)

h2(x)

E2 (kJ molÿ1)

h3(x)

E3 (kJ molÿ1)

1 2 5 10

397.5 407.9 422.7 434.1

1.0 1.0 1.0 1.0

85.6

1.0919 1.1094 1.1355 1.1565

83.5

1.1090 1.1277 1.1555 1.1779

83.4

Table 5 An iterative calculation of activation energy for dehydration of the calcium oxalate monohydrate by use of the plot of ln(b/hT2) vs. 1/T at a ˆ 0.4 b (K minÿ1)

T (K)

h1(x)

E1 (kJ molÿ1)

h2(x)

E2 (kJ molÿ1)

h3(x)

E3 (kJ molÿ1)

1 2 5 10

397.5 407.9 422.7 434.1

1.0 1.0 1.0 1.0

83.1

0.9303 0.9287 0.9264 0.9246

83.4

0.9305 0.9289 0.9266 0.9248

83.4

is good only with x ˆ 35±40, the plot of ln(b/T2) against 1/T is better when x is greater than 15. When x is lower than 15, the plot of ln(b/T2) against 1/T will also exhibit an increased error in value of calculated activation energy. Maybe the increased error in value of calculated activation energy is still smaller than the experimental error in some cases. But a perfect calculation method is always expected, at least to know how much the calculation error involved in a calculated result from the experimental data. Therefore, the iterative calculation by means of Eq. (15) or Eq. (16) is essential for obtaining the exact value of activation energy of an interested reaction, especially for the reaction of which E/RT is lower than 15. 5. Conclusions Errors in calculating the activation energy of conventional isoconversional plots, ln b versus 1/T and ln(b/T2) versus 1/T, are due to ignoring the variation of H(x) and h(x) with x. Iterative calculations based on

consideration of change in H(x) and h(x) by means of plots of ln(b/H) versus 1/T or ln(b/hT2) versus 1/T give the `exact' value of activation energy, no matter how little or how great the E/RT value of the reaction is. The `exact' value of activation energy is dependent on the `exact' p(x) adopted. The procedure of iterative calculations is general applicable, no matter what p(x) is regarded as exact p(x). The effect of the initial mass of sample on calculation of kinetic parameters should be considered. In the cases where mechanism of the being considered reaction is unknown, keeping initial mass of sample at a ®xed constant in all measurements is safe. If initial mass of sample is different in the measurements at different heating rates, the exact value of pre-exponential factor cannot be obtained at least.

References [1] T. Ozawa, Bull. Chem. Soc. Jpn. 38 (1965) 1881. [2] H.E. Kissnger, Anal. Chem. 29 (1957) 1702.

142

Z. Gao et al. / Thermochimica Acta 369 (2001) 137±142

[3] T. Akahira, T. Sunose, Trans. Joint Convention of Four Electrical Institutes, Paper No. 246, 1969 Research Report, Chiba Institute of Technology, Sci. Technol. 16 (1971) 22. [4] C.D. Doyle, J. Appl. Polym. Sci. 5 (1961) 285. [5] C.D. Doyle, Nature 207 (1965) 290. [6] A. Ortega, L.A. Perez-Maqueda, J.M. Criado, Thermochim. Acta 282/283 (1996) 29.

[7] D. Dollimore, P. Tong, K.S. Alexander, Thermochim. Acta 282/283 (1996) 13. [8] R.E. Lyon, Thermochim. Acta 297 (1997) 117. [9] R.S. Lehrle, I.W. Parsons, M. Rollinson, Polym. Degrad. Stab. 67 (2000) 21. [10] I. Amasaki, Z. Gao, M. Nakada, Chem. Lett. 5 (2000) 520. [11] C. Siat, S. Bourbigot, M.L. Bras, Polym. Degrad. Stab. 58 (1997) 303.